How is packet fragmentation accomplished in the IP standard model suite?

How is packet fragmentation accomplished in the IP standard model suite?

If an IP packet is larger than the MTU of the output interface, it will be fragmented. For efficiency purposes, the data itself is not fragmented. Instead, if a packet has to be fragmented into n pieces, n-1 dummy packets are generated and the entire data is carried in the final packet. Of course the sizes of the individual fragments are set correctly irrespective of whether or not they actually contain any data. This means that there are three kinds of IP packets:

1. Complete IP packet.
   This is a packet that has not been fragmented at all. The data filed contains the higher layer data. These packets can be identified by the fact that the data field is set.

2. Dummy Fragment.
   If an IP packet is fragmented into n pieces, the first n-1 fragments will be dummy fragments. They can be identified by the fact that neither the data nor the ip_dgram fields are set. However the contents of the IP header in the fields field and packet size will be set correctly for these packets. It is not possible to access the higher layer data from these packets.

3. Last Fragment that contains the entire IP packet.
   This last fragment would be of this type. The data field will not be set in this case also. But the ip_dgram field will contain a complete IP packet (Type 1). Even for these packets the IP header information and packet size will be set correctly.

Protocols->IP